drm/amdkfd: Disable idle optimization for chained runlist

This works around difficult-to-reproduce soft hangs on oversubscribed
runlists.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Oak Zeng <Oak.Zeng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Felix Kuehling 2019-06-20 14:54:35 -04:00 committed by Alex Deucher
parent 7a049244a0
commit a5b1615529
1 changed files with 1 additions and 0 deletions
drivers/gpu/drm/amd/amdkfd

View File

@ -134,6 +134,7 @@ static int pm_runlist_v9(struct packet_manager *pm, uint32_t *buffer,
packet->bitfields4.ib_size = ib_size_in_dwords;
packet->bitfields4.chain = chain ? 1 : 0;
packet->bitfields4.offload_polling = 0;
packet->bitfields4.chained_runlist_idle_disable = chain ? 1 : 0;
packet->bitfields4.valid = 1;
packet->bitfields4.process_cnt = concurrent_proc_cnt;
packet->ordinal2 = lower_32_bits(ib);